Our models

Grom Nova

Q: What resolution does it output?

Up to 2K natively. Drafts are best run at low resolution while you try out wordings, with the final frame taken at 2K. If you need it larger — for print, a poster or a billboard — hand it to Grom Pixel: it goes up to 16K on the long side and up to 100 megapixels.

Q: How long should the prompt be?

A long prompt is the norm for Nova, not a problem. The model holds a full paragraph and works the camera, the lens, the film stock, the light, the palette and the mood all at once. A short line works too, but the result will be more of a lottery.

Q: Which language should I write in?

It understands plain English as well as other languages. Terms about optics, film stock and genre land more precisely in English — which is why the prompts in the examples above are left exactly as they were sent.

Q: How much does one generation cost?

You pay per generation from your shared GPTunneL balance — no subscription, no monthly fee. The current price per generation is always in the price list in your account.

Q: How is Nova different from other generators on the platform?

Speed and price, with a character of its own: film texture, light and palette come to it more naturally than the sterile «perfect render». Other generators sit right next to it on GPTunneL — you can run one prompt through several and compare.

Q: Who owns the generated images?

The result is yours to use, commercial projects included. The limits are the platform's general rules and common sense: no model makes someone else's trademarks or recognisable faces legal to use.

A model for the shot that doesn't exist yet. Describe the frame in words — the optics, the film stock, the light, the palette, the mood — and get a finished image. It takes no photo on input: everything is drawn from scratch.

Open Grom Nova The whole Grom line

native resolution

16K

chained with Grom Pixel

one generation

Why

A painter, not an editor

Grom Art works on a photo you already have and changes exactly what you asked for. Grom Nova is the opposite: there is no input file at all, only your text. Everything in the frame is invented from nothing. That makes the prompt a different craft. The model cares how the shot was taken and lit: name the camera and the lens, the film stock, the light source, the palette, the mood and the genre, and it will build the frame by those rules. A long description doesn't confuse Nova — it feeds it. The more precisely you set the conditions, the closer the result lands to what you had in mind. If you already have a photo, this is the wrong model. Removing an object, extending the frame beyond its edges, recolouring clothing, restoring an old print — all of that belongs to Grom Art.

Examples

The shot and the prompt behind it

Every image was generated by Grom Nova at 2K. Click a frame to open the original — and a button to try the same prompt in the lab.

The workaround

How to feed Nova an image anyway

It takes no file itself. But Grom GPT sits right next to it — a model that sees: it looks at the photo and describes it as structured JSON. That JSON goes to Nova in place of a prompt.

Step 1

The photo

Any image — your own or one you found.

Describe image with HIGH DETAILS. All persons, objects, styles, positions. Return a BIG DETAILED structured JSON

Step 2

Grom GPT looks

One request: break the shot down in detail and return JSON.

{
  "image_description": {
    "overall_composition": {
      "subject": "A surreal and high-action scene feat…",
      "style": "Hyper-realistic photography with high…",
      "color_palette": ["Azure blue", "Bright white"…],
      "lighting": "Bright, direct sunlight creating sh…"
    },
    …

Step 3

JSON description

People, objects, style, light, palette, positions in the frame.

Step 4

Nova redraws it

It never saw the original — only the description in words.

The same chain as a pipeline: image node → Grom GPT → Grom Nova.

The chain is packaged as a Workflow template: drop in your own photo and get the result in one click.

Open the template

Bonus

And the same shot from any angle

The JSON from Grom GPT stays as it is — one line gets appended to it and the scene turns. That is how a single photo becomes a series.

Writing the prompt

Four levers Nova responds to

The model reads a description like a brief for a shoot. The more specific the conditions, the less randomness in the result.

Camera and optics

«Canon AE-1, 200mm f/2 lens» sets the framing, the compression and the depth of field far better than any amount of «make it beautiful».

Film stock and texture

Stock, grain, scan imperfections, distressed paper — this is what stops a frame from looking like a sterile render.

Light and palette

Where the light comes from, hard or soft, which colours are in play. Monochrome, two-tone, muted shadows — all of it is set with words.

Framing and mood

Close-up or wide, vertical or horizontal, genre and emotion — the model reads the orientation of the frame out of the description too.

You pay per frame, not per subscription

Each generation comes off your shared GPTunneL balance — no monthly fee, no licence, nothing to install. Trying a dozen wordings of a prompt costs less than an hour in an editor.

Pay per generation at the platform price list
One balance for Grom Nova and every other model
Low-resolution drafts cost even less

Open Grom Nova Pricing

When Nova, and when its neighbours

Choosing a model

Grom Nova

The shot doesn't exist yet: art, illustration, poster, concept. Everything is set by the description — from the lens to the palette.

Grom Art

A file on input: remove what's in the way, extend the frame, change the light and the angle, restore an old print.

Grom Pixel

The frame is headed for print or a billboard: upscaling to 100 megapixels with texture restored, not stretched.

Three ways to generate a frame

Where to try it

In the lab

Write the description and pick the result out of your generation feed. Edit the prompt and run it again.

In Workflow

A node in a pipeline: generate → pick → upscale. Handy when you need a whole series.

Over the API

The same GPTunneL key. How to connect it is in the developer documentation.

Questions about Grom Nova

/01Can I upload my own photo and rework it?

No — Grom Nova works from text only and accepts no input file. For editing an existing photo there is Grom Art: it removes objects, extends the frame beyond its edges, changes the light and the angle, and keeps the face and texture of the original. There is also a workaround: hand the photo to Grom GPT, ask it for a detailed JSON breakdown and pass that JSON to Nova — it will rebuild the frame, and can turn it to another angle while it's at it.

/02What resolution does it output?

/03How long should the prompt be?

/04Which language should I write in?

/05How much does one generation cost?

/06How is Nova different from other generators on the platform?

/07Who owns the generated images?

Everything in one GPTunneL account

The Grom models, 200+ neural networks and ready-made AI products. One balance, no subscriptions.

Open Grom Nova Get started

Grom Nova

A painter, not an editor

The shot and the prompt behind it

Analogue film

Mist and silhouettes

Grotesque portrait

Long exposure

The drawing comes alive

Sunset silhouette

Surreal scene

Stencil graphics

Paper diorama

Food advertising shot

Psychedelic poster

Mountain landscape

Architecture concept

View from the rooftop

Gothic fashion

A world in yarn

Minimalism

Hard contrast

How to feed Nova an image anyway

The photo

Grom GPT looks

JSON description

Nova redraws it

And the same shot from any angle

Front view

Camera rotated 15°

Top view

Zoom out, 30mm

Four levers Nova responds to

Camera and optics

Film stock and texture

Light and palette

Framing and mood

You pay per frame, not per subscription

When Nova, and when its neighbours

Three ways to generate a frame

Questions about Grom Nova

Everything in one GPTunneL account